AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Multi-ReduNet: Interpretable Class-Wise Decomposition of ReduNet

Fengrong Li, Delin Chu

ICLR 2026 · Architectures and efficiency › RNNs and normalization · approach: Representation geometry and collapse

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
class-wise decomposition (Multi-ReduNet) instead of ReduNetReduNet · Reuters SVMaccuracy: 0.802 -> 0.985Improved
class-wise decomposition instead of ReduNetReduNet · DrivFace and ARCENEaccuracy: 0.43-0.46 -> 0.73-1.00Improved
class-wise decomposition instead of ReduNetReduNet · MNIST SVMaccuracy: 0.906 -> 0.842Hurt
class-wise decomposition instead of ReduNetReduNet · Fashion-MNIST SVMaccuracy: 0.858 -> 0.801Hurt
add LastNorm (defer normalization to final layer) (ablation)Multi-ReduNet · six datasets best accuracybest accuracy: +0.2 pp average, +1.3 pp on Fashion-MNISTImproved
add LastNorm to reduce hyperparameter sensitivity (ablation)Multi-ReduNet · learning rate rangeaccuracy range: 9.0 pp -> 6.4 ppImproved
class-wise decomposition for training speedReduNet · six datasets training timetraining time: 1.4-2.6x speedup, about 2x averageImproved
Random Forest soft membership predictor (ablation)ReduNet · accuracy and efficiencyaccuracyNo clear effect
Multi-ReduNet-LN instead of ReduNetReduNet · Mice Protein SVM eta 0.5accuracy: 0.929 -> 0.605Hurt
Multi-ReduNet-LN instead of ReduNetReduNet · CIFAR-10 KNN eta 0.5accuracy: 0.182 -> 0.276Improved
Multi-ReduNet-LN instead of LDALDA · ARCENEaccuracy: 0.878 -> 0.829Hurt
Multi-ReduNet-LN instead of global PCA + SVMGlobal PCA + SVM · Reutersaccuracy: 97.5% -> 98.8%Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy